Academic

Measuring the Fragility of Trust: Devising Credibility Index via Explanation Stability (CIES) for Business Decision Support Systems

arXiv:2603.05024v1 Announce Type: new Abstract: Explainable Artificial Intelligence (XAI) methods (SHAP, LIME) are increasingly adopted to interpret models in high-stakes businesses. However, the credibility of these explanations, their stability under realistic data perturbations, remains unquantified. This paper introduces the Credibility Index via Explanation Stability (CIES), a mathematically grounded metric that measures how robust a model's explanations are when subject to realistic business noise. CIES captures whether the reasons behind a prediction remain consistent, not just the prediction itself. The metric employs a rank-weighted distance function that penalizes instability in the most important features disproportionately, reflecting business semantics where changes in top decision drivers are more consequential than changes in marginal features. We evaluate CIES across three datasets (customer churn, credit risk, employee attrition), four tree-based classification models

A
Alin-Gabriel Vaduva, Simona-Vasilica Oprea, Adela Bara · · 1 min read · 2 views
#cs.AI #cs.LG

arXiv:2603.05024v1 Announce Type: new Abstract: Explainable Artificial Intelligence (XAI) methods (SHAP, LIME) are increasingly adopted to interpret models in high-stakes businesses. However, the credibility of these explanations, their stability under realistic data perturbations, remains unquantified. This paper introduces the Credibility Index via Explanation Stability (CIES), a mathematically grounded metric that measures how robust a model's explanations are when subject to realistic business noise. CIES captures whether the reasons behind a prediction remain consistent, not just the prediction itself. The metric employs a rank-weighted distance function that penalizes instability in the most important features disproportionately, reflecting business semantics where changes in top decision drivers are more consequential than changes in marginal features. We evaluate CIES across three datasets (customer churn, credit risk, employee attrition), four tree-based classification models and two data balancing conditions. Results demonstrate that model complexity impacts explanation credibility, class imbalance treatment via SMOTE affects not only predictive performance but also explanation stability, and CIES provides statistically superior discriminative power compared to a uniform baseline metric (p < 0.01 in all 24 configurations). A sensitivity analysis across four noise levels confirms the robustness of the metric itself. These findings offer business practitioners a deployable "credibility warning system" for AI-driven decision support.

Executive Summary

This study proposes a novel metric, Credibility Index via Explanation Stability (CIES), to quantify the robustness of model explanations in high-stakes businesses. By employing a rank-weighted distance function, CIES evaluates the consistency of a model's explanations under realistic data perturbations, addressing the current gap in Explainable Artificial Intelligence (XAI) methods. The study demonstrates the effectiveness of CIES across three datasets and multiple models, providing business practitioners with a deployable 'credibility warning system' for AI-driven decision support. The findings highlight the impact of model complexity and class imbalance treatment on explanation credibility and stability. Ultimately, CIES offers a valuable tool for enhancing the trustworthiness of AI-driven decision-making in business applications.

Key Points

  • ▸ CIES is a novel metric for measuring the robustness of model explanations in high-stakes businesses.
  • ▸ The metric employs a rank-weighted distance function to evaluate the consistency of explanations under realistic data perturbations.
  • ▸ CIES demonstrates statistically superior discriminative power compared to a uniform baseline metric across 24 configurations.

Merits

Strength in Methodological Approach

The study's reliance on a mathematically grounded metric and rigorous evaluation across multiple datasets and models enhances its validity and generalizability.

Robustness of CIES

The sensitivity analysis across four noise levels confirms the robustness of the CIES metric, ensuring its reliability in practical applications.

Demerits

Limitation in Generalizability

The study's focus on tree-based classification models and specific datasets may limit the generalizability of CIES to other model types and domains.

Potential Overreliance on Mathematical Complexity

The use of a rank-weighted distance function may add computational complexity, potentially limiting the adoption of CIES in resource-constrained business environments.

Expert Commentary

This study makes a valuable contribution to the field of Explainable Artificial Intelligence by introducing a novel metric for measuring the robustness of model explanations. The Credibility Index via Explanation Stability (CIES) offers a practical solution for enhancing the trustworthiness of AI-driven decision-making in business applications. However, further research is needed to explore the generalizability of CIES to other model types and domains, as well as its potential adoption in resource-constrained environments. Ultimately, the study's findings highlight the importance of model interpretability and trustworthiness in AI-driven decision support systems, underscoring the need for more research in this area.

Recommendations

  • ✓ Future studies should investigate the application of CIES to other model types and domains, such as deep learning models and non-tree-based classification models.
  • ✓ Researchers should explore the development of more efficient and computationally lightweight variants of CIES, enhancing its potential adoption in resource-constrained business environments.

Sources

Related Articles